Fix some 6-1-stable tests for Ruby 2.5 and 2.6

[p8]

Summary

6-1-stable currently has failing tests: https://buildkite.com/rails/rails/builds/86353

  assert_equal "<the-name #{escaped_dangerous_chars}=\"the value\"></the-name>",
    tag.public_send(:"the-name", COMMON_DANGEROUS_CHARS => "the value")

  TagHelperTest#test_tag_builder_with_dangerous_unknown_attribute_name [/rails/actionview/test/template/tag_helper_test.rb:179]:
  --- expected
  +++ actual
  @@ -1 +1 @@
  -"<the-name _______________=\"the value\"></the-name>"
  +"<the-name>{&quot;&amp;&lt;&gt;\\&quot;&#39; %*+,/;=^|&quot;=&gt;&quot;the value&quot;}</the-name>"

The test fails because the attributes hash argument has string keys.
In Ruby 2.5 and 2.6 only Symbol keys are allowed in keyword arguments.
So the argument is seen as the content argument for the tag instead of the
attributes.

This test was introduced in 123f42a.
Running the test prior to 123f42a generates the same error.
Calling to_sym on the key fixes the test.

A simpler solution would be skipping this test altogether for Ruby 2.7, but
then we might miss future regressions.

[byroot]

Have you tried:

Suggested change

	tag.public_send(:"the-name", COMMON_DANGEROUS_CHARS => "the value")
	tag.public_send(:"the-name", { COMMON_DANGEROUS_CHARS => "the value" })

Pretty sure that will work across versions.

[p8]

That still fails:

def tag_string(name, content = nil, **options, &block)
  puts "#{name}, #{content}, #{options.inspect}"
end

tag_string('a-name', "abc" => "def")
# a-name, {"abc"=>"def"}, {}
tag_string('a-name', {"abc" => "def"})
# a-name, {"abc"=>"def"}, {}
tag_string('a-name', :abc  => "def")
# a-name, , {:abc=>"def"}

I've changed the PR to call to_sym on the key instead.

[byroot]

You're right, after digging into this myself, I realized that hash is passed to **options, so yes it must be symbols.

[p8]

Thanks! I think this should be backported to 6-0-stable as well:
https://buildkite.com/rails/rails/builds/86406#adfc3986-f771-429e-9666-059b546ecbaa

[p8]

Should I create a PR?

[byroot]

Nah don't worry, I'll cherry-pick the commit.